On Thursday 13 July 2006 08:52, luc wrote:
> Le jeudi 13 juillet 2006 à 07:16 +0200, Lourens Veen a écrit :
> >
> > Here's another. We have a number of computers running GNU/Linux in
> > our dorm, most of them (K)Ubuntu. There were some security updates
> > yesterday, and I mentioned that I'd updated my machine as well as
> > the one in our livingroom. My neighbour, who had an important
> > presentation to give that day, said he hadn't yet updated his
> > machine, preferring to do that when he didn't need it for something
> > important. The following conversation followed:
> >
> > Me: "Well, the only thing they change is a few lines of code to fix
> > the bug, so the chances of breaking something are incredibly low.
> > I've never had any problems."
> >
> > Him: "Well, last time I upgraded the kernel, the proprietary nVidia
> > driver stopped working and I was left with an unusable system."
> >
> > Me: "Aha, right. I told you you shouldn't be using evil and
> > proprietary drivers..."
>
> In this case, it´s not the problem between proprietary and openess.
> What's happend each time Apple provides a set of patches ? Most of
> the times everything runs well after the upgrade.
> Hence, in this related case, simply nVidia don´t want or cannot
> follow the kernel Linux, but apparently he´s more keen to follow
> Apple. (Not sure that, the fact that Darwin is a micro-kernel make
> all the difference)

Also, there's only one Apple, and only one MacOSX kernel. It's much 
easier for them to synchronise. Making your binary-only driver work 
with a gazillion Linux-based distributions, all of which have different 
kernel configurations, is a lot harder.

When a programme becomes free software, everyone can build (on) it, and 
it can explode into a huge amount of (slightly) different 
versions/build. Open Source solves that problem: you simply add more 
maintainers, who each maintain their own version and make sure it 
works. The problem occurs when you attempt to interface between a 
proprietary programme and a free programme. Unless there is a 
well-defined ABI, the makers of the proprietary programme have to keep 
track of everything on their own.

Another way of looking at this is by analogy: rendering pixels, like 
maintaining builds, is a parallelisable problem. Attempting to maintain 
proprietary driver builds for all of the various kernel builds out 
there is akin to rendering graphics on the CPU: it's never fast enough. 
So, OGA is to driver build maintainability what OGC is to rendering 
graphics.

Ehhm... ;-)

Lourens

Attachment: pgpLk9f18aIpl.pgp
Description: PGP signature

_______________________________________________
Open-graphics mailing list
[email protected]
http://lists.duskglow.com/mailman/listinfo/open-graphics
List service provided by Duskglow Consulting, LLC (www.duskglow.com)

Reply via email to